Adding distributed poller after existing working setup

Help
1

Hey folks…

Currently having a nice VM running a perfect and working setup of librenms since a while with no issue.

What, I would like to do and not sure how to do it easily without breaking our existing setup. I would like to install a poller at each “customers” site to receive the data on our actual setup in our office.

Our idea it’s to supply our customers with a nice little rackmount box that will do the polling and send back the data.

After reading the website about that, looks like I don’t really need a full librenms install on this particular box that will handle only polling.

So the question is, what I really need to do in our own setup to allow that and what I really need to install on the remote box to make it work.

Anyone currently have a setup like what we want to do ?

thanks

2

So long as this isn’t done over a high latency link then you need to clone the repo. Setup config.php as normal but add the additional config options for distributed polling.

You must expose memcached, mysql and rrdcached (version 1.6 or higher) from your main install unless you VPN back across (probably the recommended way here).

For each poller, configure in it’s own group and then assign the devices for that client to that group so they will only be polled locally.

3

thanks @laf

So the easiest way to deploy that it’s too have a full librenms installation with the only difference that I will use

  • MySQL
  • memcached
  • rrdcached

from the main setup.

Question about that, is it possible that the “slave” poller will “autodiscover” all devices on his side and send them back to the master and “auto-add” them on my “master” librenms ?

I think that the vpn option is the best too, I will probably setup a reverse ssh tunnel that will reach our librenms box.

4

So the pollers themselves don’t really do anything local only before pushing them to the main master poller. So if you run discovery on a remote poller and it adds a device, it adds it to the DB so it’s visible from the main poller itself.

5

@laf

Just installed a little raspberry to test at a remote site and using a small SSH vpn I’m able from the rPI to have access to our main librenms box.

Both install are OK with the validate.php

Also, the only cron I’m running on my poller1 it’s the discovery and poller…

What’s about the daily.sh ?

I imagine also, I can’t use smokeping on any of my remote site…

The only issue I seem to have now, it’s from my main librenms “web” I can’t see graph. “Error drawing graph”.

6

Resolv the issue with RRDcached!

Permission issue on the rrdcached directory!

7

You probably don’t want to auto update your installs and do them manually so only run daily.sh on the main box, disable auto updates though.

8

What is the impact of this for remote sites with what would be higher latency links?

-J

9

Slower poll times

10

@laf

How works “services” into a distributed setup ?

11

@laf is there any way to migrate a bunch of device from the command line to another poller group instead of doing it one by one using the webui ?

UPDATE: I finally change them in the DB directly and works fine :wink:

12

Question - I already have a snapshot of my Libre VM. Everything is working great on it.

If I add a remote poller and decide I do not like it, can I just revert back from snapshot on my main Libre VM and continue on?

13

You can, or you could just set it up to poll from the local machine and delete the poller again.

14

But on my main Libre VM, I would have to enable RRDCache, correct?

Once that is enabled, can I revert back if I wanted to without losing data? Or does RRDCache work just fine?

I think I may not be understanding how RRDCache works and the benefits.